Human-GenAI Interaction: What are the Risks of Calling it “Collaboration”?

In the quiet study of how people learn together, a field known as computer-supported collaborative learning, researchers have long focused on the complex dance of human minds meeting. They study how students share ideas, challenge each other's thinking, and build understanding through conversation. Central to this work is the idea that true learning happens when people are not just working side-by-side, but are deeply connected in a shared mental space where they influence one another's thoughts and feelings. For decades, the tools used to support this were simple aids: whiteboards, shared documents, or video calls that helped humans connect with other humans. But a new force has entered the room. Generative artificial intelligence, the technology behind the chatbots that can write essays, solve problems, and hold conversations, is now being used by students and teachers alike. As these tools become more capable, a natural question arises: when a human works with this machine, are they collaborating? The term "collaboration" suggests a partnership of equals, a meeting of minds where both sides contribute to a shared goal. But machines do not have minds, feelings, or intentions in the way people do. This creates a tension between the language we use to describe our new tools and the reality of how those tools actually work.

Two researchers, Gaëlle Molinari and Kristine Lund, set out to examine this tension. They asked whether calling the interaction between a person and an artificial intelligence "collaboration" is an accurate description or a misleading one. To answer this, they looked at the strict definition of collaboration used in their field, which relies on specific human qualities like shared understanding, mutual influence, and emotional connection. They then compared these human qualities against what the artificial intelligence actually does. Their investigation reveals that while the machine can be incredibly helpful, the relationship it forms with a human is fundamentally different from a partnership between two people. The authors argue that using the word "collaboration" to describe this interaction is risky because it blurs the line between human and machine, potentially changing how we value human connection and how students learn.

The researchers began by tracing the history of how we think about humans and machines working together. For a long time, scientists have imagined a future where human and computer abilities blend into a single, powerful system. Early thinkers described this as a "symbiosis," where the machine extends human power, or as a "co-evolution," where both the human and the machine learn and change together over time. In recent years, this idea has been framed as "hybrid intelligence," where humans and artificial intelligence are seen as two parts of one system, each bringing different strengths to the table. Humans offer creativity, emotional depth, and ethical judgment, while machines offer speed, pattern recognition, and vast data processing. The goal of this hybrid system is to produce results that neither could achieve alone. However, the authors point out that while this sounds like a partnership, the reality of the interaction is often one-sided. The machine does not truly "learn" from the human in the same way a person learns from another; it adjusts its output based on statistical patterns, not on a genuine understanding of the human partner's goals or feelings.

When the researchers applied the strict rules of human collaboration to these interactions, several gaps appeared. True collaboration requires that both parties understand the other's perspective and work together to build meaning. In a human group, if one person disagrees, the other feels that resistance and must adjust their thinking. This friction is essential for learning. With an artificial intelligence, however, the machine does not have a perspective to hold or a mind to change. It can generate a counter-argument, but it does not "disagree" in the human sense; it simply calculates a response that fits the conversation. The authors found that the machine cannot participate in "intersubjectivity," a state where two people share a mental space and truly understand each other's inner worlds. The machine can mimic the appearance of understanding, but it does not possess the consciousness required to truly know another person.

This distinction becomes even more critical when looking at the emotional and social side of learning. Human collaboration is deeply rooted in our bodies and our feelings. We feel empathy, we get frustrated, and we build a sense of "we-ness" through shared experiences. An artificial intelligence can be programmed to respond to emotional cues, perhaps offering a comforting word when a user seems sad. But the machine does not feel sadness itself. It cannot be affected by the human in the way a human is affected by another. The authors describe this as a profound asymmetry: the human may feel a connection, but the machine remains a tool that simulates that connection without experiencing it. This creates a risk that users, especially students, might project their own emotional needs onto the machine, believing they have a reciprocal relationship where none exists.

The consequences of calling this interaction "collaboration" are significant for education. If students believe they are working with a partner, they may start to trust the machine too much, handing over their own thinking and decision-making to it. This is known as "cognitive offloading," where a person relies on an external tool to do the mental work for them. While this can make tasks easier, the authors warn that it can lead to "cognitive surrender," where the student stops monitoring their own learning and simply accepts whatever the machine produces. Research suggests that when students rely too heavily on these tools, their critical thinking skills can weaken, and they may perform worse on tests that require deep understanding. The machine might help them finish homework faster, but it might not help them learn the material.

Furthermore, the authors argue that treating the machine as a partner changes how we view human relationships. If we get used to an artificial intelligence that is always agreeable, always available, and never truly challenges us, we might start to expect the same from our human friends and teachers. We might begin to value the smooth, frictionless interaction with a machine over the messy, difficult, but ultimately more rewarding process of working with another human. The machine can be a powerful tool, but calling it a "collaborator" obscures the fact that it is a tool. It lacks the agency, the responsibility, and the capacity for genuine mutual understanding that define human partnership.

The researchers conclude that we must be careful with our language. Using the word "collaboration" to describe human-GenAI interaction is not just a semantic error; it is a conceptual shift that could reshape our educational practices and our social expectations. By insisting on the distinction between human collaboration and human-machine interaction, we protect the unique value of human connection. We ensure that students continue to develop their own minds through the challenging, reciprocal, and deeply human process of working with others, rather than outsourcing their thinking to a machine that only pretends to be a partner. The machine is a remarkable invention, but it is not a peer, and recognizing that difference is essential for the future of learning.